India, October 15, 2026: Sthānika, the applied AI research lab of Hyderabad-based PurpleTalk group, announced that its Sieve 31B decision model has secured second place worldwide on the Jev Decision Index, an open benchmark for AI decision models hosted on Hugging Face.
Sieve 31B achieved a combined Full score of 63.8, including both public and private test suites, and ranked second on the public tests with a score of 62.5. The model outperformed international competitors such as Jev and Perplexity Decider v1.1, according to Sthānika.
Performance across model sizes
The Sieve family also includes Sieve 2B Plus, Sieve 4B, and Sieve 9B Plus, each of which placed among the top in their respective size categories on the Jev Decision Index leaderboard. Sthānika said developers can deploy smaller models on on-premise GPUs for privacy-sensitive, high-volume tasks, while larger models are available for more complex requirements.
Focus on decision-making tasks
Sieve 31B is designed as a cost-efficient decision model that can be integrated into larger AI systems, offering more reliable and inexpensive decision-making than general-purpose large language models (LLMs) for routine and high-frequency tasks. The model reads questions and available choices, then returns a defined answer with a confidence score for software to act upon, making it suitable for applications where speed, consistency, and auditability are critical.
Statements from PurpleTalk leadership
“Sieve 31B shows that a focused model fine-tuned in India can perform among the best in the world at this work. The private tests matter most to us. They come from domains outside the public suite, and Sieve generalised better than most of the field. It learned the task, not the test,” said Sridhar Muppidi, co-founder of PurpleTalk.
Bharat Lingam, CEO and co-founder of PurpleTalk, said, “We believe decision models will open up a new category of agentic systems: more reliable, more controllable and fast enough for voice, customer experience and robotics. We are publishing everything needed to reproduce our results, and we invite the research community to test it, challenge it and build on it.”
Applications and future plans
Sieve builds on Sthānika’s earlier work on compact, confidence-aware decision models, which are already used in PurpleTalk products such as Ello and Mina. The lab aims to develop agentic systems with higher reliability and low latency for sectors including voice, customer experience, and robotics.
Sthānika is an applied AI research and engineering lab focused on efficient inference, calibrated decision intelligence, and India-relevant AI. PurpleTalk, founded in 2008, is a Hyderabad-based multinational technology company with over 500 employees and a portfolio of more than 950 digital products across various sectors.
Disclaimer: This article is based on a press release provided by the concerned organisation and has been edited for style and clarity with the assistance of AI tools. BusinessFortnight does not verify, endorse, or take responsibility for the claims, figures, or statements made in the original release.
